Information Technology Reference
In-Depth Information
Noticeably, this makes component & service based systems already resemble AOSE
solutions, by favouring a component-centric rather than a global system perspective.
However, they fail to provide an adequate support for these adaptive implementations,
in terms of a common adaptation model and/or adaptation steps. Thus, the developer
has to write custom, application-specific adaptation code. This makes difficult to en-
sure that a consistent and interoperable adaptation strategy is applied throughout all the
components in one application, and also to maintain, and re-use these strategies across
multiple applications.
In contrast, AOSE provides a method of abstraction and system decomposition
based on
agentification
. This transforms a software application into a goal-oriented,
autonomous agent by building a wrapper around it so it can interoperate with the rest
of the system through standard, Agent Communication Language (ACL) interfaces and
associated coordination protocols.
However, the emphasis of AOSE and associated multiagent programming platforms
and toolkits is the coordination of large scale, deliberative multiagent systems (MASs)
while issues arising from low-level, application specific functionalities are often over-
looked. Consequently, as noted in [8], using an agent platform limit the acceptance of
mobile agents as simple programming constructs, as the programmer is forced to cen-
ter its development, its programming units, and its whole applications on the concept
of agent. Rather than a middleware-type complement to traditional (object/component-
oriented) software development, agents then become the
frontware
and require the def-
inition of complex interfaces toward the application and operating system resources.
These are the main motivations for seeking tightly integrated architectures that lever-
age the different characteristics and advantages of AOSE and CBSE. In particular, the
focus of this work is the unification between agent, component and service concepts in
a single methodology for the construction of autonomic software systems with Self-*
properties. On one hand, this work aims to define a set of re-usable, modular end ex-
tensible adaptation mechanisms for component & service-based systems. On the other
hand, the same approach will produce modular and lightweight agent systems that are
tightly integrated with mainstream component & service technology.
The remainder of the paper is organized in the following manner: In order to draw
a parallel between the CBSE and AOSE paradigms and guide the design of hybrid
CBSE/AOSE systems, Section 2 summarises the popular Belief-Desire-Intention (BDI)
agent model while Section 3 examines the most important characteristics of component
& service frameworks. Section 4 draw upon the similarities between these technologies
to translate the BDI agent model into component & service-based concepts. Section 5
introduces
Self
-OSGi, a novel agent toolkit, which is developed using the Open Service
Gateway initiative (OSGi) component & service technology [12]. Section 6 illustrates
the use of
Self
-OSGi and tests its performance with a robotic application and a dy-
namic service-selection example. Section 7 provides an overview of the most significant
agent/component integration approaches attempted in past research. Finally, Section 8
summarizes the contributions of this paper and points to some of the directions to be
explored in future research.
